Omega 58 (2016) 111–127

Contents lists available at ScienceDirect

Omega
journal homepage: www.elsevier.com/locate/omega

Correct accounting for duty drawbacks with outward and inward

processing in global production networks$
Marius Häntsch a, Arnd Huchzermeier b
a
Daimler AG – Production Strategy Locations, Mercedesstraße 138, 70327 Stuttgart, Germany
b
WHU – Otto Beisheim School of Management, Burgplatz 2, 56179 Vallendar, Germany

art ic l e i nf o a b s t r a c t

Article history: We develop a new model for the correct accounting of customs duties levied on a product. We examine
Received 20 March 2014 inward and outward processing – that is, processed components can be either imported or produced in a
Accepted 15 April 2015 foreign country – in the strategic planning of a global production network. This complex modeling
Available online 28 April 2015
problem is structured with path variables, and the duty drawbacks can be simultaneously and correctly
Keywords: entered for n production stages in m market regions (with corresponding duty regions) for all products
Duty drawbacks with a maximum n-level bill of materials. We present a case study from the automotive industry to
Production networks examine whether or not the possibility of future duty rate changes or free trade agreements, such as one
Automotive between the United States and the European Union, could affect the design of a production network and
Strategic planning
hence should be considered in strategic planning. We show that correctly accounting for duty drawbacks
can lead to changes in the global footprint of production. We also demonstrate that intercontinental
trade barriers (in the form of duties) diminish working capital and entail longer delivery routes.
Eliminating these political trade barriers could increase the returns to capital while reducing both
delivery lead times and environmental costs.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction processed by “extended workbenches” in other countries as part of

The production networks in many industries have become more
internationally linked than ever. Especially in the automotive indus-
try, global added-value activities are critical for success – as we can
see with such market leaders as VW, Toyota, BMW, and Mercedes-
Benz. Yet the establishment of production networks involves more
than following market developments; relationships, restrictions, and
promotions in international trade play a key role in selecting future
production locations [26]. For many industrialized nations and
emerging markets, the automotive industry is an important one. It
symbolizes the highest technological progress of a nation's economy,
and its high job multiplier (7.5 per job; [23]) creates steady employ-
ment in automotive manufacturing. Trade restrictions, such as high
duties on vehicles or trade agreements with selected partners, help
governments to enhance the appeal of their own production location
and flow of goods. For instance, China has high import duties on
finished vehicles even though upstream products can usually be
imported with much lower duties; moreover, if upstream products
are re-exported then duties may be waived or refunded in order to
ensure the international competitiveness of local companies. There
are also lower duties upon re-import of goods that were further
☆
This manuscript was processed by Associate Editor Kuhn.
E-mail addresses: marius.haentsch@daimler.com (M. Häntsch),
arnd.huchzermeier@whu.edu (A. Huchzermeier).
an intermediate process. Western European original equipment
manufacturers (OEMs) use such a cost advantage (e.g., from Eastern
Europe) to improve their global competitiveness [21]. The latter
measures can significantly affect the appeal of production locations
and so, in general, should be considered in the strategic planning of
such networks. Duty drawbacks are rarely addressed by optimization
models found in the literature. We know of no other research
publications about the proper accounting of duty drawbacks for
globally produced products, as when their components are (repeat-
edly) manufactured or assembled in foreign countries.
Section 2 reviews the literature on accounting for duty draw-
backs in optimization models. In the next section we discuss the
relevant fundamentals of duties and duty drawbacks. In Section 4
we develop a new model that performs that accounting in full. A
case study is presented in Section 5, where the effects of a free
trade agreement between the United States and Europe are
analyzed and we present results of a study on the ideal production
footprint. In Section 6 we summarize the results, discuss some
managerial insights, and suggest future avenues of research.
2. Survey of the literature
In this section we review previous optimization models for
global supply chains that incorporate duties and duty drawbacks.

http://dx.doi.org/10.1016/j.omega.2015.04.007
0305-0483/& 2015 Elsevier Ltd. All rights reserved.
112 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127
The optimization model of Henrich [12] has been further
developed over the years with an emphasis on the automotive
industry. This model has been updated by Fleischmann et al. [8] to
include investment decisions and has been given more flexibility
by Kauder and Meyr [18] – for example, by chaining plants (see
also [17]). Despite mentioning the importance of duty drawbacks,
Kauder and Meyr omit this factor from their model (as do
Fleischmann et al.) and thus account only for duties on imported
end products in sales regions.
A fundamental work on optimizing the supply chain network is
that of Huchzermeier and Cohen [15], which is based on Huchzer-
meier's [14] dissertation. These authors have developed a stochastic,
dynamic model that accounts for real options in the strategic plan-
ning of global supply chain networks. The model maximizes global
profit after taxes, and their theory offers a way to evaluate the risks
and opportunities of flexible production networks. With respect to
international factors, no duty drawbacks are accounted for in the
duties – see the models of Pomper [28], Cohen and Lee [6], and Canel
and Das [4]. We remark that Huchzermeier's dissertation explicitly
accounts for duty drawbacks (by way of aggregation) but does not
track the flow of goods. This limitation makes it impossible to
account correctly for inward and outward processing if (as discussed
in Section 3) intermediate components are further processed in a
foreign country. In the case of multi-sourcing options, one cannot
easily identify the duty region from which a product's respective
upstream product comes. That being said, in practice a component's
origin must be clearly marked.
Kouvelis et al. [20] present a relatively simple, multi-period,
mixed integer linear programming (MILP) model for choosing
locations in a global production network. Investment decisions
are accounted for as in Bhutta et al. [3]. The Kouvelis et al. model is
distinguished by permitting investments only in the first period;
hence that model accounts only for a two-stage production net-
work. Exogenous factors include incentives, tax breaks, and duties.
As in the models of Jacob [16] and Goh et al. [10], the customs
payments in the Kouvelis model occur irrespective of the destina-
tion country. That simplification should be critically examined
because, in fact, the duty rates on products differ by destination
and region of origin. These models also overlook duty drawbacks.
Bhutta et al. [3] have developed a MILP model for location
decisions of multinational companies. When determining the ideal
Fig. 1. Correct customs calculation based on static tariff values.
production footprint, this model accounts not only for investments
and capacity but also for such exogenous factors such as exchange
rates and customs duties. The model is demonstrated and validated
with the help of multiple scenarios – in which capacity is assumed to
be unlimited – and highly simplified case studies. Bhutta et al. do
account for customs but their model does not include duty draw-
backs. The model of Chakravarty [5] looks at the static design of
supply chain networks. That model accounts for investment decisions
and variable costs; it also includes such global parameters as local
taxes, exchange rates, and customs duties. Sales prices are not
specifically addressed either in this model or in the Bhutta et al. [3]
model, and customs owed are simply added to the unit costs.
Häntsch and Huchzermeier [11] take a similar approach, assessing
the risks in strategic network planning by way of a model that
includes customs in the relevant transport costs.
Summaries of different models for planning supply chain networks
can be found in Aikens [1], Vidal and Goetschalckx [31], Erengüç et al.
[7], Goetschalckx and Fleischmann [9], Papageorgiou [27], and Melo
et al. [24]. Aikens presents, as do Goetschalckx and Fleischmann, one
of the first overviews of fundamental location decision models. In their
detailed review, Vidal and Goetschalckx classify the various models in
terms of the factors for which they account. Erengüç et al. analyze
possible planning approaches for operational, tactical, and strategic
supply chain planning; Goetschalckx and Fleischmann focus on the
strategic level. Melo et al. give a broad and detailed review of facility
location models in the context of supply chain management. Although
that review considers duties, it merely conflates duties, exchange rates,
and some other financial aspects as “internal factors”; thus, duties are
mentioned only as an aside.
In short, only a few authors account for duty drawbacks in their
models. One explanation for this reluctance is that correctly calculat-
ing the route of materials or upstream products requires one to trace
them in full from the source to the end customer. Tracing a path
increases the complexity of a model's formulation and practically
assures a longer run time for the optimization model.
We will use the example that follows to establish the relevance
of duty drawbacks (see Fig. 1). We will demonstrate how the
models that account for duties – but not for duty drawbacks – can
result in significant discrepancies when calculating netduties. A
cost calculation that fails to account for duty drawbacks results in
customs duties of €200 for the import of the powertrain to a
 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127
NAFTA country (2.5%n€8000) and customs duties of €4000 for
each vehicle imported from a NAFTA region into the EU (10%n
€40,000). Thus, a total of €4200 in customs is charged. However,
in reality, it is only €3200 ((40,000
8000)n10%) because the EU
views this as outward processing. Thus the value subject to
customs is reduced by €8000. In this case, the difference in the
duty costs is 25%.
Arntzen et al. [2] developed one of the earliest models (for a
computer company) that accounts for many factors in the interna-
tional supply chain, including duty drawbacks. The authors note that
this model enabled the company to reconfigure its supply chain and
to achieve $100M (US) in savings ([2], p. 69). Their model is
considered to be innovative for this reason and others (cf. [19], p.
84), in part because it is one of the few models that incorporates
throughput time in the objective function. However, the procedure
for optimizing overall throughput time is problematic because it can
result in longer delivery times for some customers ([19], p. 84). The
Arntzen et al. model lacks several important details. Kohler notes that
upstream products from suppliers are not considered in duty draw-
backs. Moreover, both Hübner [13, p. 61] and Kohler [19, p. 136]
observe that the mathematical model's description of duty draw-
backs does not explain how they are implemented in practice.
Neither is a numerical solution given for any case study.
Smith [29] develops a model for a tier-1 supplier in the auto-
motive industry. This model optimizes the last production stage in
the internal production network, for which the optimal supply base
should be chosen. The production plants of the automotive company
taking delivery are clearly defined, as are the locations of own (tier-2)
suppliers. Because all products studied are within the same tariff
zone, when transported on a connection they are assigned the same
duty rate. Smith's model also greatly simplifies the notion of duty
drawbacks, which is fortunate because a precise formulation would
result in a nonlinear optimization model.
Meyer [25] presents a model for selecting the optimal location
that incorporates (among other factors) duty drawbacks. However,
the author greatly simplifies some aspects of the problem, which
makes one wonder how closely the model reflects reality [19, p.
136]. For instance, Meyer does not account for product structure or
materials flow and does not view products in factories separately
[13, p. 83].
Oh and Karimi [26] have designed their model especially for the
process industry, so it allows in particular for divergent product
structures. However, that model does not account for duty drawbacks
with outward processing, and its limitation to a single production
step restricts one's options when using it. Hübner [13] also developed
a model with the process industry in mind; his model likewise
supports divergent product structures. Yet unlike Oh and Karimi,
Hübner does not limit the number of production stages. Note that
this latter model does not account for outward processing, either. The
author remarks that the model can in general properly account for
inward processing if the product imported into a country is re-
exported after just one processing step.
Kohler [19] presents the most extensive model to date that
incorporates duty drawbacks. He uses a multi-objective function
to optimize net cash flow and minimize throughput time. This
mixed integer model can be used to decide between different
manufacturing process or open locations, for instance. However,
Kohler can present duty drawbacks correctly only by limiting the
material flow's flexibility so much that, for each market and
product, only one factory can perform the same production step.
This approach excludes, for example, tandem production of a
product at multiple locations for the same market.
In their paper, Mariel and Minner [22] present a model for
calculating duties and duty drawbacks for unlimited stages, sourcing
options, products, and markets; this model also exploits product
flexibility at each stage. The authors demonstrate their model using a
113
solution method that they developed for a select case study. To track
the exact flow of products in the supply chain, Mariel and Minner use
the concept of production configurations (rather than a path variable
approach). Hence all information about the regions of pre-products is
given by the so-called bill of materials, which leads to total
enumeration of paths. This is a smart and easy way to track the
exact product flow for smaller production networks, but it does have
limitations. If there are large numbers of regions, sites, pre-products,
and product stages, then the number of bills of materials that the
planner must consider (and solve) will increase exponentially.
It can fairly be said that – with the exception of Mariel and
Minner [22] – models in the literature either do not properly
account for duty drawbacks with inward and outward processing,
calculate (import) duties on only an aggregated basis, or control
material flow tightly with just a few stages in a convergent supply
chain. It should also be noted that (static) tariff values, and not the
actual value added, are the basis for any correct calculation and
crediting of drawbacks. This topic is of increasing importance
because customs authorities are now more likely to focus on the
traceability of each individual (main) component.
Hence this paper's contribution to the field lies in enabling parallel
production of a product at multiple locations for the same market
while correctly accounting for duties, including their drawbacks.
3. Fundamentals of duty drawbacks and challenges for the
optimization model
3.1. Inward processing
Inward processing offers a way to waive duties on goods
imported from outside the duty region if those goods are exported
in the same or processed form [30, p. 223]. Some of the literature
refers to “export in same condition” [2, p. 76] for unaltered goods
and “export in different condition” for altered goods (p. 77). The
European Union (EU) provides for both a suspension system and a
duty drawback system as follows [30, p. 223].
Suspension system: Goods can be imported without import duties.
Duty drawback system: When goods are imported, the regular
import duties are to be charged first. However, if the imported goods
are re-exported at a later stage then the import duties are refunded
upon application. The process is illustrated in Fig. 2. A product – in
this case, an engine – is imported into the United States, where it is
processed further (i.e., installed in a vehicle during the assembly
process). The vehicle is then exported. Because the engine underwent
further processing in the United States but the processed product
(the vehicle) was then exported and left the country, we have inward
processing. In this case, either method (suspension or duty draw-
back) can be used to forgo import duties on the engine.
3.2. Outward processing
Outward processing is the converse of inward processing. In
outward processing, products from the respective duty region –
which are temporarily exported before they are subject to another
value-adding process in a foreign country – are assessed with a
lower duty upon re-import of the processed goods [30, p. 207].
This scenario is illustrated in Fig. 3 using an auto powertrain as an
example. Here the value of the vehicle subject to duty, which is
being imported into the EU, is reduced by the value of the
powertrain, which also comes from the EU. The usual method
(since 1 July 2001) for determining the tariff value is the added
value method. The processing costs are used as the tariff value
subject to duty. This corresponds to the added value that occurred
in the foreign country [30, p. 213].
114 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127

Fig. 2. Duty drawback under inward processing.

Fig. 3. Reduction in value subject to duty under outward processing.

Note that the difference between the tariff value at level i and
the tariff value of level i
1 may not be equal to the actual value
added at level i. Throughout the text we assume that the tariff
value is static, since dynamic changes are seldom accepted by int-
ernational customs agencies.
3.3. Challenges for presentation in the optimization model
Next we identify the requirements for an optimization model
capable of capturing inward and outward processing.
Requirements for inward processing. Rules for inward processing
state that a material imported into a country is not subject to
customs if it will be re-exported, whether or not it undergoes a
processing step. In terms of our model, this means that we must
know whether imported materials will remain in the country to
which they were exported. There are, in particular, two complica-
tions worth noting. First, we must remember that even for multiple
production stages followed by (re-)export, no import duties are to
be charged; see Fig. 4. Hence we must expressly track the material
flow from import to export. Note also that, for interim re-export to a
third country, again no duties are to be paid; see Fig. 4. The ori-
ginally imported material (here, a camshaft) is used in an engine in
the country of final processing. This engine will be exported for a
further processing step in a foreign country, so there are no customs
due on the import of that camshaft. Custom payments – perhaps
accounting for any discounts associated with outward processing –
will be necessary if the material, following its later re-import (here,
in the vehicle end product), is to remain permanently in the EU or a
third country.
Requirements for outward processing. The requirements for out-
ward processing place additional demands on the optimization
model to be created. This will be clarified with an example in
Fig. 5. If a product is re-imported following temporary export after
a processing step, then only the value added outside the EU is
subject to customs; see Fig. 5. Conversely, value that was added in
the EU can be deducted from the transfer price. The added value of
 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127 115

Fig. 4. Examples of inward processing with re-import.

Fig. 5. Examples of outward processing with one and multiple processing steps.

upstream products is therefore credited with this duty rate upon
re-import with the processed good. If we compare this sequence
with re-import after multiple processing steps, as in Fig. 5b, then
the added value can be given as a reason to reduce the customs for
the same materials upon import of the vehicle. Yet because a
higher duty rate applies to the final product, the added value of
the upstream product is credited at 10% (EU import duty). For the
model developed here, the implication is that we must know not
only the route a component travels but also the product in which it
will ultimately be imported into a country.
4. Development of the model
To meet the model requirements just described, in this section
we develop a model that uses a solution based on path variables.
In a path variable solution, a variable is defined for each
component of a bill of materials (see Fig. 6) that has no prior
component(s) as inputs (i.e., for each “basic component”). This
variable clearly shows the flow of that component all the way to
reaching a customer. A basic component typically can have multi-
ple routings, which lead to additional decision variables in
the model.
Fig. 6 offers an example with five locations and three duty
regions. For one unit of the end product, which consists of two
basic components, we have the following manufacturing options:
path variable III or IV for component (part) p21; and path variable
I, II, or V for component p22. For example, one solution is p21 via
III and p22 via I. In all, there are three combinations that will
satisfy the expected demand (assuming that the basic component
p22 from factory f23 cannot be delivered to factory f11).
Of course, the cost of using path variables is increased complex-
ity, but this is the only way to calculate duties correctly. The
complexity stems mainly from a product's number of basic com-
116 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127

Fig. 6. Sample definition of path variables I–V based on a given supply chain structure and bill of material (a) Supply chain for product p1 in duty from a2. and (b) Bill of
material of product p1.

ponents (i.e., the end nodes in a bill of materials), not from the rDR set of all resources R
number of intermediate products. U D P  F set of all nodes in the production network
Our model will use a two-level bill of materials for a two-stage Ui D U set of all nodes in stage i
production process. (We assume that there is a match between pðuÞ product p of node u ¼ ðp; f Þ A U
levels and stages.) The factories and markets are assigned to the f ðuÞ factory f of node u ¼ ðp; f Þ A U
duty region in which they are located. Calculation of duties for an V D P  Aset of nodes in the production network for all duty
n-stage production process is explained in Section 4.3.2. regions
Vi DV set of all nodes in stage i for all duty regions
4.1. Objective function pðvÞ product p of node v ¼ ðp; aÞ A V
aðvÞ duty region a of node v ¼ ðp; aÞ A V
The model's objective function minimizes the discounted cash Fa set of the factories located in duty region a
flow CF t of the production network. Here CF t is discounted by the ViðaÞ D Vi set of all nodes in stage i in duty region a A A
weighted average cost of capital WACC for the first planning iðvÞ D Vi stage i of node v A V
period FP while accounting for possible lead times in planning. Y u2;u1;u0;t path variable Y indicating the quantity of products p2
8 9 that is manufactured in a stage-2 factory for an end
<X =
CF t product p1, which is produced in f1 and flows to market
min ð1Þ
: t ð1 þ WACCÞt
FP ; f0 in period t
X v2;v1;v0;t path variable X indicating the quantity of products p2
The cash flow per period t is based on Eq. (2). It consists of that is manufactured at stage 2 in region a for an end
production costs Prodt (Eq. (12)), transport costs Transt (13), inc- product p1, which is produced at stage 1 in region a for a
entives Incent t (14), investment costs Invest t (15), and duty costs market in region a0 in period t
Dutyt (10). Note that Dutyt is the total of all duties for all products #
in all duty regions. Shortfalls are penalized in the optimization for The optimization model is based on three sets of decision
each period t with the cash flow Shortf allt (16). variables: variables for location decisions f actOpenft 1 , for the
f 1;r
CF t ¼ Prodt þ Transt
Incent t þ Invest t þ Shortf allt þDutyt ð2Þ resource decisions of a location resAvailt , and for material flow
Y u2;u1;u0;t .
The binary variable f actOpenft 1 gives the model the freedom to
4.2. Decision variables decide whether a location for final assembly f A F1 should be open
in period t (see 4.9 and 4.10 for further model details). A second
In the following part of the paper we use set variables to enh- f 1;r
binary variable, resAvailt , gives the model the freedom to decide
ance the readability of the model. if a production plant f A F1 has a resource r A R available in period t
Definitions (see 4.9 and 4.10). To facilitate discussion of the model, we set
u2 A U2, u1 A U1, and u0 A U0 as tuples of a factory at stage i for
a index of duty regions product pi.
i index of production stages The decision variable Y u2;u1;u0;t captures the permissible flow of
A set of duty regions a basic component from the lowest stage u2 via u1 to u0 in period
F set of all factories t. Because only the path via each duty region is relevant for our
f i D Fi set of all factories in stage i (i ¼ 0 : marketÞ duty calculations, the flow is derived from those regions. The
f ia D Fia set of all factories in stage i (i ¼ 0 : marketÞ of a region a variable X v2;v1;v0;t gives the flow from v2 via v1 to v0 in period t
P set of all products P with v2 A V2, v1 A V1, and v0 A V0. With this variable X v2;v1;v0;t
pi D Pi set of all products or components of a stage i which is derived from Y u2;u1;u0;t the flow of a basic component is
p1 D P1 set of all end products fully described by the individual duty regions. In particular, the
 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127
flow of all (components or) end products from stage 1, X v1;v0;t can
be expressly defined using v1 A V1 to stage v0 A V0 along with
the bill of materials. The difference between the two path variables
Y u2;u1;u0;t and X v2;v1;v0;t is merely that the information about
factories is replaced by information about the relevant duty
regions.
4.3. Duty costs (Duty)
This section will omit the period index t in order to improve
readability.
Definitions
Duty sum of all duty costs for all regions
DUTYRATEv1;v0 duty rate between region aðv1Þ and aðv0Þ for a
product p1ðv1Þ
DutyRegv0 total duty costs in region aðv0Þ
REGVALADDv2 regional value added for a product p2ðv2Þ in region
aðv2Þ
TARIFFV ALv1;v0 tariff value of product pðv1Þ from a region aðv1Þ to
region aðv0Þ
TARIFFV ALv2;v1 tariff value of product pðv2Þ from a region aðv2Þ to
region aðv1Þ
tvv1;v0 cumulative tariff value of the entire transported volume
of a product pðv1Þ on the way from region aðv1Þ to
market region aðv0Þ
tvv2;v0 cumulative tariff value of the entire transported volume
of a product pðv2Þ on the route from duty regions aðv2Þ to
aðv0Þ
tvCredit1_2v1;v0 credit for inward processing, based on regional
value added at production stage 2
4.3.1. Two-stage supply chain model with duty drawbacks
The principal aim of this paper is calculating the duty to be charged
while accounting for duty drawbacks. In models with two production
stages and any number of duty and market regions, the goal is to
define value-added credits for which we can account in an end
product's duty (see Fig. 7a). This means that the credit for the added
value from the same duty region is deducted from the tariff value of
the end product. In that case, the tariff value of the upstream products
cannot be used because prior value-added activities may be included
in the tariff value. Recall that the tariff value is used only for calculating
the duty payment to be charged upon crossing the border.
This approach is taken separately for each path variable. The
correct duty costs are calculated in three steps while accounting
for inward and outward processing:
1. reduction of the tariff value of a product through outward
processing;
2. tariff value subject to duty; and
3. duty costs.
These three steps are explained next.
Step 1: Reduction of the tariff value of a product through
outward processing
We start by determining the value of regional value added. Eq. (3)
defines value added at production stage 2, which tra-
nspired in the duty region of the sales area. With outward pro-
cessing, components are exported and re-imported; note that the
117
export and re-import of components is no longer possible at stage 1.
X v0 A V 0;
tvCredit1_2v1;v0 ¼ REGVALADDv2 nX v2;v1;v0
v2 A V2ðaðv2Þ ¼ aðvoÞÞ
v1 A V1
ð3Þ
Step 2: Value subject to duty
Eq. (4) calculates the tariff value of the entire transported
volume of a product pðv1Þ on the way from region aðv1Þ to
market region aðv0Þ: First, the tariff value TARIFFVALv1;v0 of
product pðv1Þ is multiplied by X v1;v0 . This figure is reduced by
the credit for inward processing, which is based on the regional
value added at production stage 2: tvCredit1_2v1;v0 .
v0 A V0;
tvv1;v0 ¼ TARIFFVALv1;v0 nX v1;v0
tvCredit1_2v1;v0 ð4Þ
v1 A V1
Eq. (5) uses the same procedure for the entire transported
volume of a product pðv2Þ on the route from duty regions aðv2Þ
to aðv0Þ. Because duty is owed only if the product does not
leave the duty region again in further production steps leading
up to the market, only the “downstream” nodes v1 that are
located in duty region aðv0Þ are added up.
X v0 A V0;
tvv2;v0 ¼ TARIFFVALv2;v1 nX v2;v1;v0 ð5Þ
v1 A V1ðaðv1Þ ¼ aðv0ÞÞ
v2 A V2
Step 3: Duty costs
The duty costs are calculated from Eq. (6) using the tariff value
and duty rate for each stage. First, the duty costs are calculated for
the transport of all products p1 on the route between the duty
region aðv1Þ and the market region aðv0Þ. The cumulative tariff
value tvv1;v0 for the entire transported volume of a product on a
route is multiplied by the duty rate DUTYRATEv1;v0 of the
corresponding product. The total of all products and all routes
from aðv1Þ to aðv0Þ results in the entire duty to be charged for
stage 1; the same procedure is then followed for stage 2. Eq. (7)
summarizes the duty costs for all regions.
X
DutyReg v0 ¼ tvv1;v0 nDUTYRATEv1;v0
v1 A V1
X
þ tvv2;v0 nDUTYRATEv2;v0 v0 A V 0 ð6Þ
v2 A V2
X
Duty ¼ DutyReg v0 ð7Þ
v0 A V 0
See Appendix A for an example that uses real-life numbers
(derived from Fig. 2) for the duty cost calculation.
4.3.2. n-stage supply chain model with duty drawbacks
An n-stage model can be described as follows. Decision variable
X vn;…;v0;t contains the volume of a basic component pðvÞ of stage n
from duty region aðvnÞ, which flows via the route vn; …; v0 to the
market region aðv0Þ in period t.
X vn;…;v0;t vn A V n; …; v A V0; 8 t
The intermediate path variables for components and end
products vn
1; …; v0 (and so forth) can be derived directly from
the bill of materials for any n r 1.
Step 1: Reduce the value of a product subject to duty p through
outward processing
Eq. (8) shows the value of regional added value that can be
deducted because of outward processing. For each path variable,
the tariff value of the last transfer into the market region is used
118 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127

Fig. 7. Value subject to duty at production stage 1 and 2.

Fig. 8. (a) Definition of value added to be credited for an imported end product and (b) definition of the relevant tariff value in general.

to define the duty. Only the value added that is “upstream” of the Step 3: Duty costs
transfer can be credited; see Fig. 8. The duty costs are based on Eqs. (10) and (11). Eq. (10) defines
the duty costs per region, and (11) defines the total duty costs
for all regions (and products).
X
n X
tvCredit_i_nvi;…;vo ¼
j ¼ iþ1 vj A Vj Xn X
DutyReg v0 ¼ tvvi;vo nDUTYRATEvi;vi
1 vO A V 0 ð10Þ
vi A Vi; …;
ðaðvjÞ ¼ aðv0ÞÞREGVALLADDvj nX vj;…;vi þ 1;vi;…;vo ð8Þ X i ¼ 1 v A Vi
vO A V0 Duty ¼ DutyReg v0 ð11Þ
v0 A V 0
Step 2: Value subject to duty
Eq. (9) calculates the tariff value to be considered for the entire
transported volume of a product pðviÞ on the route from region The following list gives the terms (and their definitions) used in
aðviÞ to region aðv0Þ. Then tvvi;v0 implies that the product the model's formulation of a two-stage supply chain.
will not leave the respective region – that is, because aðvi

1Þ ¼ aðvi
2Þ ¼ … ¼ aðv0Þ.
X BIGM big number M#
tvvi;vo ¼ TARIFFVALv;vi
1 nX vi;…;vo BOM1p2;p1 bill of materials 1 – required product volume p2 to
vi
1 A Vi
1ðaðvi
1Þ ¼ aðv0ÞÞ
vi A Vi manufacture a product p1

tvCredit_i_nvi;…;vo ð9Þ
v0 A V0 CAPACITY1ft 1 capacity of the factory f 1 in period t
CAPACITY2ft 2 capacity of the factory f 2 in period t
 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127 119

CAPAREQ 1fp1;t
1
capacity utilization for the production of a product 4.6. Incentives Incentt
p1 in factory f 1 in period t
Eq. (14) calculates the incentives1 that are given either as a
CAPAREQ 2fp2;t
2
capacity utilization for the production of a product fixed incentive for the factory or as a variable incentive for each
p2 in factory f 2 in period t product manufactured in that factory. So incentives are added up
f0
Demandp1;t demand for product p1 in markets f 0 during period t either for all factories or for all products produced in a factory.
X XX
INCENTFIX1ft 1 fixed incentives for a factory f 1 in period t Incent t ¼ INCENTFIX1ft 1 þ INCENTV AR1fp1;t
1
nY u1;u0;t
INCENTFIX1ft 2 fixed incentives for a factory f 2 in period t X
F1
XX
F1 P1

INCENTVAR1fp1;t 1
variable incentive for product p1 manufactured þ INCENTFIX1ft 2 þ INCENTVAR2fp2;t
2
nY u2;u1;u0;t 8t
in factory f 1 in period t F2 F2 P2

INCENTVAR2fp2;t 2
variable incentive for product p2 manufactured ð14Þ
in factory f 2 in period t 4.7. Investment costs Investt
INVESTF ft 1 cost to open a factory f 1 in period t
INVESTRft 1;r investment costs for a resource r in factory f 1 in Investment costs are incurred, for instance, when a new factory
period t is opened at stage 1 or when the firm invests in new resources for
Incent t incentives in period t one of these factories. The investment costs are ultimately calcu-
Invest t investment costs in period t lated according to (15) using the total costs for opening factories
MINLOADft 1 minimum utilization per factory f 1 per period t plus the total of all investment costs for resources in level-1
PRODFIX1ft 1 fixed costs of a factory f 1 in period t factories.
PRODFIX2ft 2 fixed costs of a factory f 2 in period t X XX
Invest t ¼ INVESTF ft 1 nopenNewF ft 1 þ INVESTRft 1;r nopenNewRft 1;r 8t
PRODVAR1fp1;t 1
variable production costs for a product p1 produced F1 F1 R
in a factory f 1 in period t ð15Þ
PRODVAR2fp2;t 2
variable production costs for a product p2 produced
in a factory f 2 in period t 4.8. Cost of shortfalls Shortfallt
Prodt production costs in period t
Shortf allt costs of shortfalls in period t To ensure that the demand is met in all markets, an optimiza-
TRANS1fp1;t
1;f 0
transport costs for the transport of a product p1 from tion plan must calculate the cost of shortfalls. The procedure inv-
a factory f 1 to a market f 0 in period t olves multiplying a big number BIGM by the shortfall, which con-
TRANS2fp2;t
2;f 1
transport costs for the transport of a product p2 from sists of demand minus the delivered volume of all factories f 1 per
a factory f 2 to a factory f 1 in period t market f 0 of product p1.
!
Transt transport costs in period t XX f0
X
openNewF ft 1 binary (indicator) variable set equal to 1 only if the Shortf allt ¼ BIGM n Demandp1;t
Y u1;u0;t 8t ð16Þ
F0 P1 P1
factory f 1 is newly opened in period t
openNewRft 1;r binary (indicator) variable set equal to 1 only if the
resource r of a factory f 1 is newly opened in period t 4.9. Other decision variables
openein period t installed in period t
Two auxiliary variables are used for the structure of the
production network. Eq. (17) indicates the period t in which a
production facility f 1 will be opened. In period t Z FP, an opening
4.4. Production costs Prodt
takes place if a factory is open (f actOpenft 1 ¼ 1) and it is not an
existing factory (ISNEWF f 1 ). For all other periods, the equation
Production costs are those costs that are directly assignable to the
indicates whether a factory is open in period t yet was not open in
production processes at a particular location. The costs comprise the
the prior period (t
1).2
total of all fixed costs PRODFIX ft 1 and PRODFIX2ft 2 of the production 8
plants at both levels, PRODFIX ft 1 and PRODFIX2ft 2 , and from the total < f actOpenf 1 nISNEWF f 1 if t Z FP
f1 t
of the variable production costs PRODVAR1fp1;t1
and PRODVAR2fp2;t 2
for openNewF t ¼ 8 f 1; t
: f actOpenft 1
f actOpenft 1
1 otherwise
each product at both levels – each multiplied by the volume of a
product produced in the factory. We remark that no fixed costs are ð17Þ
incurred at stage 1 unless a production factory is actually open. Eq. (18) determines in which period t a resource r1 is made
X XX
Prodt ¼ PRODFIX1ft 1 nf actOpenft 1 þ PRODVAR1fp1;t1
nY u1;u0;t available at a factory f 1. In period t Z FP, an installation takes place
f 1;r1
F1 F1 P1 if a resource is available (resAvailt ) and the resource was pre-
f 1;r1
X XX viously not available (ISNEWR ¼ 1). For all other periods, (18)
þ PRODFIX2ft 2 þ PRODVAR2fp2;t
2
nY u2;u1;u0;t 8t determines whether a factory is open in period t yet was not open
F2 F2 P2 in period t
1.3
ð12Þ 8
4.5. Transport costs Transt < resAvailf 1;r1 nISNEWRf 1;r1 if t Z FP
f 1;r1 t
openNewRt ¼ 8 f 1; r1; t
: resAvailft 1;r1
resAvailf 1;r1 otherwise
t
1
The transport costs are calculated by (13) using either the
transport costs for a product from one factory f 2 to another factory ð18Þ
f 1 or the transport costs for a product p1 from factory f 1 to market
f 0 (TRANS1fp1;t
1;f 0
Þ. These transport costs are multiplied by the trans-
1
ported volume and then summed for all products and routes. Incentives are offered by governments for the construction of plants, or on a per-
product basis, so that the country will seem more attractive as a business location.
XX XX 2
Eq. (17) also ensures that a factory cannot be closed, which means that the
Transt ¼ TRANS1fp1;t
1;f 0
nY u1;u0;t þ TRANS2fp2;t
2;f 1
nY u2;u1;u0;t 8t equation cannot have a negative evaluation.
F1 P1 F2 P2 3
Eq. (18) also ensures that a resource cannot be removed; this means that it
ð13Þ cannot have a negative evaluation, either.
120 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127

4.10. Conditions f 1;r

resAvailt A ½0; 1 8 f 1; t; r ð29Þ

The following condition ensures that all factory-f2 deliveries of

the upstream product p2correspond to the actual need for this
product in the factory f1 for the production of the product p1 for 5. Case studies
market f0.
X In this section we present two case studies. The first demon-
Y u2;u1;u0;t ¼ Y u1;u0;t nBOM1p2;p1 ð19Þ strates the effects of correctly calculating duty drawbacks, includ-
F2
ing effects on the monetary result, materials flow, and production
Condition (20) ensures that no more products p1 are delivered structure. This case study is based on one production network of a
to a market f0 than are demanded. German automotive manufacturer. All reported quantitative fig-
X ures are based on true values.
Y u1;u0;t r DEMANDfp1;t
0
ð20Þ
The second case study demonstrates the extent to which the
F1
possibility of a future free trade agreement (FTA) between the United
The following equations ensure compliance with respect to the States and the EU could affect the design of a production network.
capacity limits of factories in level 1 and level 2. Condition (21) This study also indicates, for a firm engaged in strategic planning,
ensures compliance with the capacity of factory f1. The total of how best to account for such possible changes in duty rates.
factory f1's manufactured products p1, when multiplied by its The results will be presented and interpreted after we explain
individual capacity utilization CAPAREQ 1fp1;t1
, must be smaller in all the structure of these case studies.
periods t than the factory's capacity. Multiplication by the binary
decision-making variable f actOpenft 1 also ensures that, for a non-
5.1. Structure of the case studies
open factory, the capacity limit is zero. Condition (22) ensures that
a product cannot be manufactured unless the required resource
f 1;r1 Products and product structure. The products shown here cover
(RESREQ r1
p1 ) is in place – that is, unless resAvailt ¼ 1.
XX part of the portfolio of an automotive company that uses the same
Y u1;u0;t nCAPAREQ 1p1;t r CAPACITY1t nf actOpenft 1 8 t ð21Þ
f1 f1
factories and logistical structures. Products in our model are five
F1 P1 vehicle types that are equipped with five different engines and a
XX f 1;r1
set of upstream products for each vehicle type. The vehicle types
Y u1;u0;t nCAPAREQ 1fp1;t
1
r CAPACITY1ft 1 nresAvailt 8t ð22Þ are mid-size A (MidA), mid-size B (MidB), luxury (Lux), sports
F1 P1
utility vehicle (SUV), and roadster (Roadster). Engine types include
Condition (23) ensures compliance with the capacity limits for gasoline versions with either four cylinders (EngG4) or six cylin-
any factory f2. ders (EngG6) as well as diesel versions with either four cylinders
XX (EngD4) or six cylinders (EngD6). Another engine module covers
Y u2;u1;u0;t nCAPAREQ 2fp2;t
2
r CAPACITY2ft 2 8 t ð23Þ
all other engines. To simplify matters we assume that the engine
F2 P2
modules are identical for all five vehicle types. Finally, a set of
Condition (24) requires minimum utilization for a factory f1 in parts for each model (PartsMidA, PartsMidB, PartsLux, PartsSUV,
period t. and PartsRoadster) will come as upstream products from Europe.
XX
Y u1;u0;t nCAPAREQ 1fp1;t
1
Z MINLOADft 1 8 t ð24Þ Markets. Ten markets are studied. In addition to the traditional
F1 P1 triad markets (United States, Europe, and Japan) we have the
growth markets of BRICS (Brazil, Russia, India, China, and South
Condition (25) ensures that a stage-1 factory f1 is open in period t
Africa), which exhibited increasing sales figures and high duty
if it was opened in the prior period t
1. This secondary condit-
rates during the period of study. Mexico was also added to the list.
ion rules out closing locations once they’ve been opened. Moreover,
All other markets are summarized as “rest of world” (RoW).
a factory that was not marked “new” (i.e., a factory for which
Production locations and suppliers. Five existing assembly plants
ISNEWF f 1 ¼ 0) should be open from the first period onward.
8 and five possible future plants are depicted in Fig. 9. The existing
< ð1
f actOpenf 1 ÞnðISNEWF f 1
1Þ if t ¼ FP plants are in Germany (DE1 and DE2), the United States (US),
t
0r 8 f 1; t ð25Þ South Africa (ZA), and China (CN). New factories could be built in
: f actOpenft 1
f actOpenft 1
1 otherwise
Mexico (new_MX), Brazil (new_BR), India (new_IN), and China
Condition (26) ensures that a resource r1 of a level-1 produc- (new_CN2. New factories cannot be opened until 2015, and each
tion network f 1 in period t is available if it was available during would require an investment of €1.1B. Production locations for the
the prior period t
1. This condition makes it impossible to close engine as an upstream product, hereinafter “engine factories”, are
locations that have already been opened. Analogously to (25), a in China (EngCN), the United States (EngUS), and Germany
factory resource that is not marked as new (ISNEWRf 1;r1 ¼ 0) (EngDE). One European supplier (PartsEU) delivers to all factories.
should be available from the first period onward. Resources. All existing factories have been provided only cer-
8 tain resources to manufacture vehicles thus far. For strategic
< ð1
resAvailf 1 ÞnðISNEWRf 1;r
1Þ if t ¼ FP reasons, in the future they should continue to invest only in
t
0r 8 f 1; r; t
: resAvailft 1
resAvailf 1 otherwise resources for select vehicle categories. New factories should also
t
1
invest only in select resources (see Table 1). Investment in new
ð26Þ resources is possible from the first period onward and amounts
Condition (27) ensures that a resource is available to a factory to €200M.
only if that factory is open. Capacity. Each assembly plant has a specified maximum capa-
f 1;r
city – based on factory bottlenecks – to which it must adhere. The
f actOpenft 1 Z resAvailt 8 f 1; r; t ð27Þ sites DE1 and DE2 produce (respectively) 500,000 and 400,000
Finally, our binary variables must also satisfy the following units annually. The US site produces 300,000 units each year; CN
conditions. and ZA produce 300,000 and 150,000, respectively. All future
factories are to have a capacity of 300,000 production units per
f actOpenft 1 A ½0; 1 8 f 1; t ð28Þ year. A minimum utilization rate of 60% is required for existing
 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127 121

Fig. 9. Global production footprint with existing and potential new factories.

Table 1 Use of duty drawbacks with inward and outward processing. We

Existing and possible resources for each factory. assume that the firm will exploit all the advantages of inward and
outward processing that can be achieved through an enhanced
Production plant Existing resources Possible resources
network structure and optimized product flow – whether or not
DE1 MidA, Lux, Roadster MidA, MidB, Lux, Roadster
duty drawbacks are accounted for when the network is optimized.
DE2 MidA, MidB, Roadster MidA, MidB, Lux, Roadster Production structure. Products can be delivered from any factory
US MidA, SUV MidA, MidB, SUV to any market. Exceptions are CN and new_CN2, which can supply
CN MidA MidA, MidB, SUV only the Chinese market. Engines can be delivered from EngDE to
ZA _ MidA, MidB
any assembly plant. Engines from EngUS can be delivered only to
new_TR _ MidA, MidB
new_MX _ MidA, MidB US, new_MX, and new_BR; engines from EngCN, only to CN and
new_BR _ MidA, MidB new_CN2.
new_CN2 _ MidA, MidB, SUV Changes in duties. Any changes in duties should be fully
new_IN _ MidA, MidB
implemented at the start of 2015, though the pending changes
should be known already in 2013. Hence changes to the network
structure (e.g., building new factories) can be planned and imple-
mented in advance.
factories. The capacity of engine factories in China and the United
States is limited; in order to avoid bottlenecks in the model, we
5.2. First case study
assume that both the capacity of the German engine factory and
the delivery of upstream products are unlimited.
In this section we present and interpret the results of our first
Tariff value. The tariff value is based on the value of the vehicle,
case study. We describe the network structure and product flow
the transport costs, and the insurance premium (until it crosses
first without and then with the incorporation of duty drawbacks.
the border for import).
Afterward, we compare the monetary values under the two
Duty rate. The duty rates for finished vehicles are shown in
approaches. To solve the case study we use the XPRESS solver on
Table 2. For engines, duty rates for the Harmonized System (HS)
a standard lap top computer (with an Intel Core i5 CPU unit
code number4 8407.34 (2500-cc gasoline engine) are used. For the
operating at speed of 2.4 GHz with 4 GB RAM). The solution time
upstream product parts, an average duty rate is used for the HS
was less than one minute.
number 8708.29 (parts and accessories for vehicles in positions
8701–8705) for each import country; see Table 3.
Credit for outward processing. The value credited for outward 5.2.1. Optimization without application of duty drawbacks
processing for an upstream product corresponds to the cost of First we present the results under the optimized network
production or the purchase price of the relevant upstream product. structure, after which we describe the capacity utilization of each
Observation period and discount. A ten-year (2013–2022) plan- factory and then investigate product flow properties. The changes
ning horizon is used. The cash flow is discounted to the year 2013 to the network structure are shown in Fig. 10b. Except for ZA, the
at the rate of 10%. existing production networks do not produce new vehicle classes;
ZA will begin producing MidB starting in 2022.
Three new factories will be opened. In 2015, the new_MX
4
These HS numbers are assigned by the United Nations to classify goods for factory will be opened to produce MidA and MidB. One period
customs purposes. later, in 2016, a factory will be opened in India (new_IN) that will
122 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127
Table 2
Import duties for vehicles.
also produce MidA and MidB. In 2018, another factory will open in
China (new_CN2); this factory will first produce MidB and later,
starting in 2022, also MidA. Factory utilization is tabulated in
Fig. 10a. In the existing factories DE1, DE2, and ZA, only the
prescribed minimum utilization applies (except in the first two
periods). The US factory will increase its capacity to range between
66% and 84%. The factory in China will increase capacity in later
periods and will be at full utilization in 2017. Utilization of the new
factories will vary greatly. The factory in Mexico (new_MX) will
start at 50% utilization and increase to 88% utilization by 2022. The
additional factory in China (new_CN2) will start in 2018 at 19%
utilization and increase continually to 63% in 2022. The lowest
utilization rate is that of the new factory in India (new_IN), which
will not exceed 16% in any period.
Next we analyze the product flows of vehicle categories MidA
and MidB. Product flows of the other vehicle categories are already
prescribed by definition; namely, the Lux and the Roadster must
be produced in Germany, and SUVs must be US-built for all
markets except China. The factories in China supply only the
Chinese market and so these, too, are excluded. The production
of MidA and MidB vehicles, which exceeds the minimum utiliza-
tion of DE1 and DE2, will be moved to new_MX once that location
opens, at which time all US MidA production will also move there.
This new location will primarily serve the markets of Brazil, Japan,
Mexico, and the United States as well as (starting in 2017) the RoW
countries. After the Indian factory opens in 2016, the production of
MidA and MidB vehicles will move from DE1 and DE2 to India.
However, new_IN will produce vehicles for the Indian market only.
Note also that engines for SUVs produced in the United States for
the EU market come from the EngUS factory. This fact is significant
when we compare, in Section 5.2.3, the results from optimization
that incorporates duty drawbacks (Section 5.2.2).
5.2.2. Optimization with application of duty drawbacks
As in Section 5.2.1, we first present the results under the
optimized network structure; we then discuss factory utilization
and product flow properties. The analysis is again restricted to the
MidA and MidB vehicle segments.
The network structure is summarized in Fig. 11b. Of the existing
factories, ZA will also start making MidB vehicles starting in the
first period of 2013. Three new factories will be opened. In 2015
the new_IN factory will begin producing MidA and MidB vehicles,
as will the new_MX factory in 2019; in 2018, the new_CN2 factory
will be opened to manufacture MidB vehicles.
With regard to factory utilization, Fig. 11a shows that DE1 and
DE2 meet only the minimum level. The ZA factory is full in the first
two periods, but it falls to the minimum level once the new Indian
factory is opened. The US site operates at a utilization rate ranging
from 74% to 84% – enough to meet the entire world's demand (not
including China) for SUVs. The newly opened factories will also
have different utilization. Whereas the factory in India will seldom
fall below half capacity (46–57%) upon opening in 2015, its
capacity will drop precipitously (to 11–16%) when the Mexican
factory opens in 2019. This is because the production volume
initially relocated from South Africa to India in 2015 will be
transferred to Mexico starting in 2019. The new_MX factory will
be at 67% utilization when it opens, increasing to 88% within three
years (by 2022).
Product flow explains these varying rates of utilization. Whereas
the ZA factory will initially produce MidA and MidB for its own
market and all MidB vehicles for the US market, its production
volume will change in 2015 upon the opening of new_IN because
that location will then handle all production for its own market. By
2018, it will also supply the RoW export market in full and a large
part of the Japanese market for MidA and MidB vehicles. Thus,
 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127 123

new_IN will take over part of the production volume of DE1 and (without and with the application of duty drawbacks), we are
DE2. So in order to meet their minimum utilization levels, the latter finally in a position to compare them. After doing so, we shall
two factories must take over production volume of both vehicle interpret the differences.
categories for the US market from the ZA factory. In 2017, Europe
will witness a small jump in demand for Lux and Roadster vehicles
as well as a larger increase in demand for MidA vehicles – all of 5.2.3. Comparison of results
which will be produced in DE1 and DE2. Therefore, production Network structure, factory utilization, and product flow. In both
volumes of MidA and MidB for the US market, which exceed the solutions, a new factory will be opened in Mexico, China, and
minimum utilization of DE1 and DE2, can again be relocated to ZA India. The new_CN2 factory will be opened once the existing
and thereby quickly increase its utilization. This product flow will factory in China (CN) reaches the limit of its capacity. It will serve
first change with the opening of new_MX in 2019, at which time only the Chinese market and so will not be further considered.
that location will take over all production volume from new_IN However, the order in which new_MX and new_IN open differs:
excepting that designated for the Indian market. Another key factor the new_MX factory is the later-opening one when duty draw-
is the origin of upstream products. If duty drawbacks are applied to backs are applied. Especially in view of ZA's utilization in the first
the engines of US-made SUVs for the EU market, then those engines two periods, the application of duty drawbacks results in produc-
must come not from EngUS but rather from EngEU. tion volume being relocated more frequently from DE1 and DE2 to
Having presented the main optimization results for production distant locations with lower production costs. Another difference
structure, factory utilization, and product flows for both scenarios when duty drawbacks are considered is that new_IN, at least for a
time, will produce vehicles for export markets.
Table 3 The relocation of production volume in the first two periods to
Import duties for engines and parts. ZA shows that import duties for upstream products strongly affect
the appeal of ZA as a producer of export vehicles. Applying inward
From To
processing thus enhances the appeal of producing export vehicles
EU USA Turkey Brazil India China Mexico South Africa in factories that must first import most of their upstream products.
In this way, inward processing encourages remote production.
HS 8407.34 gasoline engine 2500 cc Optimization without applying duty drawbacks also has the
EU 0 2.5 0 18 7.5 10 0 0
new_IN factory being opened mainly in response to the high import
USA 2.7 0 2.7 18 7.5 10 0 0
China 2.7 2.5 2.7 18 7.5 0 0 0 duties on vehicles destined for the Indian market. The overhead to
produce vehicles for export is higher than that of a new_MX factory.
HS 8708.29 product parts
EU 0 2.5 0 18 10 10 0 15
Applying the duty drawbacks due to inward processing reveals that
it can be economical to use the Indian factory for producing export

Fig. 10. Results without the application of duty drawbacks.

124 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127
Fig. 11. Results with the application of duty drawbacks.
vehicles as well. Therefore, the construction of a new factory will be
delayed by four years. Observe that the potential factories in Turkey
(new_TR) and Brazil (new_BR) will not be opened in any event.
Although the new_TR factory would have less production overhead
than the existing factories in DE1 and DE2, its costs for duties and
vehicle transport are similar to those for DE1 and DE2. As a result,
the required minimum utilization of existing factories renders
another factory in Europe an unattractive option.
Of course, the new_BR factory is always less attractive than
new_MX given the former's small local market and higher duties
for exporting vehicles, especially to the United States.
Finally, consider the origin of the engines for the SUVs
produced by the US factory for the EU market. As noted previously,
these engines come from the EngEU factory – despite the higher
transport costs – when duty drawbacks are taken into account.
The reason is that lower duties can be achieved in the EU market if
SUVs are imported from the US factory. These savings in duties
exceed the engine's initially higher transport costs.
Monetary volume. We now examine the monetary effects. Table 4
shows the change in discounted cash flow that results from the
proper application of duty drawbacks. It is worth emphasizing that,
when calculating these values, we assume that the production
network resulting from the optimization will be perfectly managed
(i.e., will reflect the application of all possible duty drawbacks). Here
the net financial value added when duty drawbacks are applied is
about €464M. The largest positive influence comes from production
(€491M). Advantages from duties (€365M), investments (€210M),
and incentives (€83M) also contribute. (The savings from inward
processing for SUVs from the US for the EU market amount to some
€208M.) The savings in investments result, for example, from the
delay in opening new_MX factory. Against these positive values are
the higher expenses for transport (€685M) when duty drawbacks are
incorporated. The expense of transporting upstream products
accounts for most (€599M) of that negative value.
5.3. Second case study: US–EU free trade agreement
Our second case study looks at the effect of duty rate changes on a
production network – in particular, the effects of a possible free trade
zone between the United States and the EU. For the purpose of this
simulation we assume that the FTA is implemented at the start of year
2015, whereupon no more duties are owed between these two trade
zones); we also assume that, in 2013, all parties become aware of the
pending changes. This means that required changes to the network
structure, such as the construction of new factories and allocation of
vehicles, can be planned and executed in advance. We shall describe
how this FTA changes not only the network structure and product flow
but also the discounted cash flow; we then interpret these findings.
Network structure, factory utilization, and product flow. Fig. 12b
shows the structure of our production network after a US–EU free
trade zone is introduced. We shall compare this structure with
that resulting from the basic scenario (i.e., correct application of
duty drawbacks; see Fig. 11).
The only change in the network structure is that the factory in
new_MX will be opened one year later. The utilization of ZA, DE2, and
new_IN will change in 2017 and 2018 because the production volumes
of MidA and MidB will be allocated differently among the various
factories. During those years, DE1 and DE2 (rather than ZA) will
produce additional vehicles in both segments for the US market. That
allows for outsourcing of more production volume for the Japan and
RoW markets to new_IN, which is why capacity there will increase.
The one-year delay in opening new_MX translates into increasing the
2019 utilization of both DE2 and new_IN. The US factory will use only
US engines, even in SUVs produced for the EU market. This is because
SUVs can be imported without duty into the EU thanks to the free
trade agreement. Therefore, engines from the EU can no longer be
used to reduce the duties paid.
Monetary volume. Table 5 shows that a free trade zone between
the United States and the EU leads to better monetary results, as
 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127
detailed in the next paragraph. Production costs will rise, mainly
because the delayed opening of new_MX requires that the 2019
production volume be manufactured in factories with relatively
higher production costs.
The largest part of the savings in this scenario is directly related
to reduced duties, in particular those on SUVs built at the US
factory for the EU market. Vehicles from the DE1 and DE2 factories
for the US market likewise increase savings. The duties for
upstream products are also reduced because they are now
imported duty-free by the United States from the EU. Transport
costs will be reduced, mostly as a result of lower transport costs
for upstream products. A sizable share of these cost reductions is
due to eliminating the transport of engines from the EU to the
United States; in the absence of an FTA, those EU engines were
formerly installed in US-made SUVs destined for the EU market.
There is also a positive change for the firm's investments, since an
extra year will elapse before the Mexican factory opens.
Table 4
Monetary effects of accounting for duty drawbacks.
Costs ΔDCF 2013–2022 [M€]
Production
Duty
thereof for vehicles þ400
thereof by SUV from US for EU þ 331
thereof for upstream products
35
Transport
thereof for vehicles
86
thereof for upstream products
599
thereof engines for SUV for EU
105
Incentives
Investments
New locations and allocation
Total:
Fig. 12. Results under a US–EU free trade agreement.
125
Altogether, the net improvement in discounted cash flow is
€2341B. This amount reflects the elimination of EU import duties
on US SUVs and the related elimination of transport costs of
engines from the EU, a difference of €1989B (or 85% of the total
savings). All other changes from the optimization (as compared
with the basic scenario) yield a further advantage of some €352M.
The results presented here suggest that a free trade zone offers a
large savings potential. The main effect results from SUVs being
manufactured only in the US factory. An additional effect – resulting
from the delayed opening of the factory in Mexico – is much less
significant. Hence the previously optimized production network (i.e.,
sans an FTA) is also beneficial. The only adjustment of consequence in
this scenario is the outsourcing to EngUS of production volume for
(upstream) engines to be used in US-made SUVs destined for Europe.
That outsourcing would amount to 65,000 engines per year. A free
trade zone would make it preferable to produce upstream products
in the duty region of the assembly factory.
In contrast, the rules on outward processing make it preferable to
produce upstream products in the duty region of the sales market.
Thus, the free trade zone tends to outsource production of the end
product to a region with low production costs and also tends to
produce upstream products in that region. As a result, lead times are
þ491 shorter and – thanks to shorter transport routes – total carbon dioxide
(CO2) emissions are reduced. In light of the monetary and production
þ 365
footprint effects demonstrated previously, the possibility of changes in
duty rates (e.g., a US–EU FTA) should be accounted for when firms
engage in the strategic planning of production networks.

685
6. Summary and outlook
þ 83 We have shown that correct application of duty drawbacks has
þ 210
a large effect on network structure under inward and outward
þ 464 processing. Duty drawbacks due to inward processing make the
manufacture of export vehicles outside the EU more attractive (the
126 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127

Table 5 Appendix A
Monetary effects of a US–EU free trade agreement.
Assume a three-stage supply chain. The company ships a powertrain
Costs ΔDCF 2013–2022 [M€]
ðp2Þ from the EU ðaðv2ÞÞ to the United States ðaðv1ÞÞ. In the United
Production
34 States, the powertrain is built into a vehicle ðp1Þ that is then shipped back
Duty þ2119
to the EU ðaðv0ÞÞ. According to our model, there is no duty associated
thereof for vehicles þ2005 with the EU powertrain. The following data is used in this example:
thereof by SUV from US for EU þ 1625
X v2;v1;v0 ¼ 1
thereof from Lux/Roadster for US þ 364
thereof from other vehicles þ 16 REGVALADDv2 ¼ €8000
thereof for upstream products þ 114 TARIFFVALv1;v0 ¼ €40; 000
thereof from upstream products for US þ103
TARIFFVALv2;v1 ¼ €8000
Transport þ152
thereof for vehicles
10 DUTYRATEv1;v0 ¼ 10%
 
thereof for upstream products þ 162 DUTYRATEv2;v0 ¼ 0% because av2 ¼ av0
thereof for engines to US from EU þ105

Incentives
13
Investments þ 117
New locations and allocation Step 1: Reduction of the tariff value of a product through
Total: þ 2341 outward processing
X v0 A V0;
tvCredit1_2v1;v0 ¼ REGV ALADDv2 nX v2;v1;v0
v2 A V 2ðaðv2Þ ¼ aðvoÞÞ
v1 A V1

higher the location's import duties, the greater the effect). The ðB:1Þ
correct application of duty drawbacks due to outward processing
tvCredit1_2v1;v0 ¼ €8000n1 ¼ €8000
results in upstream products being produced more frequently in
the region of the respective vehicle factory. The duty drawbacks
associated with outward processing also influence the network Step 2: Value subject to duty
structure. Outsourced production that does (resp., does not) v0 A V0;
account for that effect will prefer to source upstream products in tvv1;v0 ¼ TARIFFVALv1;v0 nX v1;v0
tvCredit1_2v1;v0
v1 A V1
the duty region of the factory (resp., the market).
ðB:2Þ
The results of our first case study show that correct application
of duty drawbacks can compensate for longer transport routes. tvv1;v0 ¼ €40; 000n1
€8000 ¼ €32; 000
However, the resulting positive monetary effects are accompanied
by negative effects – for instance, increased environmental pollu- X v0 A V0;
tion due to higher CO2 emissions from those longer transport tvv2;v0 ¼ TARIFFVALv2;v1 nX v2;v1;v0
v1 A V 1ðaðv1Þ ¼ aðv0ÞÞ
v2 A V2
routes, which in turn result in longer lead times and higher capital
commitment costs. Longer routes also make the firm more ðB:3Þ
vulnerable to fluctuating logistics costs and increase the difficulty tvv2;v0 ¼ 0 ½because aðv1Þ a aðv0Þ
of resolving quality defects, especially since errors might not be
identified until the engine has been installed and since replace-
ment parts may be a long time coming. Another consequence of Step 3: Duty costs
X
longer transport routes is increased risks of route disruption, DutyReg v0 ¼ tvv1;v0 nDUTYRATEv1;v0
damage, and theft while the goods are being transported. v1 A V 1
X
Upstream products would have to be manufactured sooner if þ tvv2;v0 nDUTYRATEv2;v0 v0 A V0 ðB:4Þ
outward processing were applied, since longer transport times v2 A V 2
would be required before final assembly. In the case study, the
engines (for SUVs made in the United States for export to Europe) X
had to be completed nearly three months before final assembly Duty ¼ DutyReg v0 ðB:5Þ
v0 A V 0
simply to accommodate the delivery time from Europe to Amerca.
DutyReg v0 ¼ €32; 000n10%þ 0 ¼ €3200 Duty ¼ €3200
In such circumstances, demand must be estimated as accurately as
possible in order to avoid bottlenecks, delayed delivery, and excess
production. This estimating requirement could increase the firm's
vulnerability to a “whiplash” effect. To confirm our findings, References
future research might employ multi-criteria optimization models.
Another topic worth exploring is the extent to which “flexibility [1] Aikens CH. Facility location model for distribution planning. European Journal
strategies” could accommodate changes in duty rates. of Operational Research 1985;22:263–79.
The optimization model presented in this paper allows one to [2] Arntzen B, Brown G, Harrison T. Global supply chain management at digital
equipment corporation. Interfaces, vol. 25; 1995.
account for duty drawbacks explicitly. It contributes to the literature [3] Bhutta K, Huq F, Frazier G, Mahamed Z. An integrated location, production,
on global supply chain modeling by showing how to improve the distribution and investment model for a multinational corporation. Interna-
optimizing of production networks while taking consistent account tional Journal of Production Economics 2003;86:201–16.
[4] Canel C, Das S. Modeling global facility location decisions: integrating market-
of duty drawbacks.
ing and manufacturing decisions. Industrial Management and Data Systems
In conclusion, the model is simply a tool that can be used to 2002;102(2):110–8.
improve our understanding of a global production network's com- [5] Chakravarty A. Global plant capacity and product allocation with pricing
plexity. One should bear in mind that the final user (e.g., top decisions. European Journal of Operational Research 2005;165(1):157–81.
[6] Cohen M, Lee H. Resource deployment analysis of global manufacturing and
management) is responsible for determining which aspects of the distribution networks. Journal of Manufacturing and Operations Management
model are useful and how its results should be interpreted. 1989:81–104.
 M. Häntsch, A. Huchzermeier / Omega 58 (2016) 111–127
[7] Erengüç Ş, Simpson N, Vakharia A. Integrated production/distribution plan-
ning in supply chains: an invited review. European Journal of Operational
Research 1999;115:219–36.
[8] Fleischmann B, Ferber S, Henrich P. Strategic planning of BMW's global
production network. Interfaces 2006;36(3):194–208.
[9] Goetschalckx, M, Fleischmann B. Supply chain management and advanced
planning. In: Stadtler H (Hrsg.), Kilger C (Hrsg.), Strategic network design.
Heidelberg: Springer-Verlag; 2008.
[10] Goh M, Lim J, Meng F. A stochastic model for risk management in global supply
chain networks. European Journal of Operational Research 2007;182:164–73.
[11] Häntsch M, Huchzermeier A. Identifying, analyzing, and assessing risk in the
strategic planning of a production network: the practical view of a German car
manufacturer. Journal of Management Control 2013:125–58.
[12] Henrich P. Strategische Gestaltung von Produktionssystemen in der Auto-
mobilindustrie. Dissertation. Augsburg: University of Augsburg; 2002.
[13] Hübner R. Strategic supply chain management in process industries an
application to specialty chemicals production network design. Lecture notes
in economics and mathematical systems. Berlin: Springer-Verlag; 2007.
[14] Huchzermeier A. Global manufacturing strategy planning under exchange rate
uncertainty. Dissertation. University of Pennsylvania; 1991.
[15] Huchzermeier A, Cohen M. Valuing operational flexibility under exchange rate
risk. Operations Research 1996;44(1):100–13.
[16] Jacob F. Quantitative Optimierung dynamischer Produktionsnetzwerke. 2006.
[17] Jordan W, Graves S. Principles on the benefits of manufacturing process
fexibility. Management Science 1995;4(41):577–94.
[18] Kauder S, Meyr H. Strategic network planning for an international automotive
manufacturer. OR Spectrum 2009;31(3):507–32.
[19] Kohler K. Global supply chain design. Dissertation. Konzeption und Implemen-
tierung eines multikriteriellen Optimierungsmodells für die Gestaltung globaler
Wertschöpfungsaktivitäten. Julius-Maximilians-Universität Würzburg; 2008.
127
[20] Kouvelis P, Rosenblatt M, Munson C. A mathematical programming model for global
plant location problems: analysis and insights. IIE Transactions 2010;36:127–44.
[21] Loch CH, Chick S, Huchzermeier A. Management quality and competitiveness:
lessons from the industrial excellence award. Springer; 2008 156 p.
[22] Mariel K, Minner S.). Strategic capacity planning in automotive production
networks under duties and duty drawbacks. In: Grubbström RW, Hinterhuber
HH, editors. Pre-prints of the 18th international working seminar on produc-
tion economics, vol. 4, Innsbruck; 2014. p. 129–42.
[23] McAlinden S., Fulton G. Contribution of the Automotive Industry to the U.S.
Economy; 1998.
[24] Melo MT, Nickel S, Saldanha-da-Gama F. Facility location and supply chain
management – a review. European Journal of Operational Research 2009;196
(2):401–12.
[25] Meyer T. Globale Produktionsnetzwerke – ein Modell zur kostenoptimierten
Standortauswahl. Magisterarbeit: Universität Aachen; 2005.
[26] Oh H-C, Karimi I. Global multiproduct production – distribution planning with
duty drawbacks. AIChE Journal 2006;52(2):595–610.
[27] Papageorgiou L. Supply chain optimization for the process industries:
advances and opportunities. Computers and Chemical Engineering
2009;33:1931–8.
[28] Pomper C. International investment planning: an integrated approach.
Amsterdam: North-Holland Publishing Company; 1976.
[29] Smith R. The value of real options fexibility demonstrated in the global
automotive industry. Global supply chain performance and risk optimization.
Gabler Edition Wissenschaft, Deutscher Universitäts-Verlag; 2002.
[30] Thoma A, Böhm R, Kirchhainer E. Zoll und Umsatzsteuer – Die rechtliche
Beurteilung und praktische Abwicklung von Warenlieferungen mit Drittlands-
bezug. 2nd ed.Springer Gabler; 2010.
[31] Vidal CJ, Goetschalckx M. Strategic production-distribution models: a critical
review with emphasis on global supply chain models review. European
Journal of Operational Research 1997;98(1):1–18.